undecidability in group theory, topology, and...
TRANSCRIPT
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Undecidability in group theory, topology, andanalysis
Bjorn Poonen
Rademacher Lecture 2November 7, 2017
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Group theory
Question
Can a computer decide whethertwo given elements of a group are equal?
To make sense of this question, we must specify
1. how the group is described
2. how the element is described
The descriptions should be suitable for input into a Turingmachine.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Group theory
Question
Can a computer decide whethertwo given elements of a group are equala given element of a group equals the identity?
To make sense of this question, we must specify
1. how the group is described
2. how the element is described
The descriptions should be suitable for input into a Turingmachine.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Group theory
Question
Can a computer decide whethertwo given elements of a group are equala given element of a group equals the identity?
To make sense of this question, we must specify
1. how the group is described
2. how the element is described
The descriptions should be suitable for input into a Turingmachine.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Group theory
Question
Can a computer decide whethertwo given elements of a group are equala given element of a group equals the identity?
To make sense of this question, we must specify
1. how the group is described: f.p. group
2. how the element is described: word
The descriptions should be suitable for input into a Turingmachine.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Example: The symmetric group S3
In cycle notation, r = (123) and t = (12). These satisfy
r3 = 1, t2 = 1, trt−1 = r−1
It turns out that r and t generate S3, and every relationinvolving them is a consequence of the relations above:
S3 = 〈r , t | r3 = 1, t2 = 1, trt−1 = r−1〉.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Finitely presented groups
Definition
An f.p. group is a group specified by finitely manygenerators and finitely many relations.
Example
Z× Z = 〈a, b | ab = ba〉
Example
The free group on 2 (noncommuting) generators is
F2 := 〈a, b | 〉
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Representing elements of an f.p. group: words
S3 = 〈r , t | r3 = 1, t2 = 1, trt−1 = r−1〉.
Definition
A word is a sequence of the generator symbols and theirinverses, such as
tr−1ttrt−1rrr .
Since r and t generate S3, every element of S3 is representedby a word, but not necessarily in a unique way.
Example
The words tr and r−1t both represent (23).
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
The word problem
Given an f.p. group G , we have
Word problem for G
Find an algorithm with
input: a word w in the generators of Goutput: YES or NO, according to whether w = 1 in G .
Harder problem:
Uniform word problem
Find an algorithm with
input: an f.p. group G , and a word w in thegenerators of G
output: YES or NO, according to whether w = 1 in G .
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Word problem for Fn
Theorem
The word problem for the free group Fn is decidable.
Algorithm to decide whether a given word w represents 1:
1. Repeatedly cancel adjacent inverses until there isnothing left to cancel.
2. Check if the end result is the empty word.
Example
In the free group F2 = 〈a, b〉, given the word
aba−1bb−1abb,
cancellation leads toabbb,
which is not the empty word,so aba−1bb−1abb does not represent the identity.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Undecidability of the word problem
Theorem (P. S. Novikov and Boone,independently in the 1950s)
There exists an f.p. group G such that the word problem forG is undecidable.
The strategy of the proof, as for Hilbert’s tenth problem, isto build a group G such that solving the word problem for Gis at least as hard as solving the halting problem.
Corollary
The uniform word problem is undecidable.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Markov properties
Definition
A property of f.p. groups is called a Markov property if
1. there exists an f.p. group G1 with the property, and
2. there exists an f.p. group G2 that cannot be embeddedin any f.p. group with the property.
Example
The property of being finite is a Markov property, because
1. There exists a finite group!
2. Z cannot be embedded in any finite group.
Other Markov properties: trivial, abelian, free, . . . .
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Theorem (Adian & Rabin 1955–1958)
For each Markov property P, the problem of decidingwhether an arbitrary f.p. group has P is undecidable.
Sketch of proof.
Embed the uniform word problem in this P problem:Given an f.p. group G and a word w in its generators,build another f.p. group K such that
K has P ⇐⇒ w = 1 in G .
Example
There is no algorithm to decide whether an f.p. group istrivial.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Fundamental groupFix a manifold M.
and a point p.Consider paths in M that start and end at p.Paths are homotopic if one can be deformed to the other.
Fundamental group π1(M) := {paths}/homotopy.
Group law: concatenation of paths.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Fundamental groupFix a manifold M and a point p.
Consider paths in M that start and end at p.Paths are homotopic if one can be deformed to the other.
Fundamental group π1(M) := {paths}/homotopy.
Group law: concatenation of paths.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Fundamental groupFix a manifold M and a point p.Consider paths in M that start and end at p.
Paths are homotopic if one can be deformed to the other.
Fundamental group π1(M) := {paths}/homotopy.
Group law: concatenation of paths.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Fundamental groupFix a manifold M and a point p.Consider paths in M that start and end at p.Paths are homotopic if one can be deformed to the other.
Fundamental group π1(M) := {paths}/homotopy.
Group law: concatenation of paths.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Fundamental groupFix a manifold M and a point p.Consider paths in M that start and end at p.Paths are homotopic if one can be deformed to the other.
Fundamental group π1(M) := {paths}/homotopy.
Group law: concatenation of paths.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Examples of fundamental groups
π1(torus) = Z× Z π1(sphere) = {1}
This gives one way to prove that the torus and the sphereare not homeomorphic, i.e., that they do not have the sameshape even after stretching.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
The homeomorphism problem
Question
Given two manifolds, can one decide whether they arehomeomorphic?
To make sense of this question, we must specify how amanifold is described.
This will be done using the notion of simplicial complex.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Simplicial complexes
Definition
Roughly speaking, a finite simplicial complex is a finite unionof simplices (points, segments, triangles, tetrahedra, . . . )together with data on how they are glued. The description ispurely combinatorial.
Example
The icosahedron is a finite simplicial complexhomeomorphic to the 2-sphere S2.
From now on, manifold means “compact manifoldrepresented by a particular finite simplicial complex”,so that it can be the input to a Turing machine.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Undecidability of the homeomorphism problem
Theorem (Markov 1958)
The problem of deciding whether two manifolds arehomeomorphic is undecidable.
Sketch of proof.
Let n ≥ 5. Given an f.p. group G and a word w in itsgenerators, one can construct a n-manifold ΣG ,w such that
1. If w = 1 in G , then ΣG ,w ≈ Sn.
2. If w 6= 1, then π1(ΣG ,w ) is nontrivial (so ΣG ,w 6≈ Sn).
Thus, if the homeomorphism problem were decidable, thenthe uniform word problem would be too. But it isn’t.
In fact, the homeomorphism problem is known to be
decidable in dimensions ≤ 3, and
undecidable in dimensions ≥ 4.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
The previous proof showed that for n ≥ 5, the manifold Sn isunrecognizable: the problem of deciding whether a givenn-manifold is homeomorphic to Sn is undecidable.
Theorem (S. P. Novikov 1974)
Each n-manifold M with n ≥ 5 is unrecognizable.
Question
Is S4 recognizable? (The answer is not known.)
To explain the idea of the proof of the theorem, we need thenotion of connected sum.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Connected sumThe connected sum of n-manifolds M and N is then-manifold obtained by cutting a small disk out of each andconnecting them with a tube.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Am I a manifold?
Theorem
It is impossible to decide whether a finite simplicial complexis homeomorphic to a manifold.
Proof.
SΣG ,w := suspension over our possibly fake sphere ΣG ,w .
If w = 1 in G , then ΣG ,w ≈ Sn, so SΣG ,w ≈ Sn+1.
If w 6= 1, then SΣG ,w is not locally Euclidean.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Knot theory
Definition
A knot is an embedding of the circle S1 in R3.
Definition
Two knots are equivalent if one can be deformed into theother within R3, without crossing itself.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
From now on, knot means “a knot obtained by connecting afinite sequence of points in Q3”, so that it admits a finitedescription.
Theorem (Haken 1961 and Hemion 1979)
There is an algorithm that takes as input two knots in R3
and decides whether they are equivalent.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Higher-dimensional knots
Though the knot equivalence problem is decidable, ahigher-dimensional analogue is not:
Theorem (Nabutovsky & Weinberger 1996)
If n ≥ 3, the problem of deciding whether two embeddings ofSn in Rn+2 are equivalent is undecidable.
Question
What about n = 2? Not known.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities
Question
Which of the following inequalities are true for all real valuesof the variables?
a2 + b2 ≥ 2ab
TRUE
x4 − 4x + 5 ≥ 0
TRUE
536x287196896 − 210y287196896 + 777x3y16z4732987
−1111x54987896 − 2823y927396 + 27x94572y9927z999
−936718x726896 + 887236y726896 − 9x24572y7827z13
+89790876x26896 + 30y26896 + 987x245y6z6876
+9823709709790790x28 − 1987y28 + 1467890461986x2y6z4
+80398600x2z12 − 27980186xy + 3789720156y2 + 9328769x
−1956820y − 275893249827098790768645846898z ≥ −389?
FALSE
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities
Question
Which of the following inequalities are true for all real valuesof the variables?
a2 + b2 ≥ 2ab TRUE
x4 − 4x + 5 ≥ 0
TRUE
536x287196896 − 210y287196896 + 777x3y16z4732987
−1111x54987896 − 2823y927396 + 27x94572y9927z999
−936718x726896 + 887236y726896 − 9x24572y7827z13
+89790876x26896 + 30y26896 + 987x245y6z6876
+9823709709790790x28 − 1987y28 + 1467890461986x2y6z4
+80398600x2z12 − 27980186xy + 3789720156y2 + 9328769x
−1956820y − 275893249827098790768645846898z ≥ −389?
FALSE
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities
Question
Which of the following inequalities are true for all real valuesof the variables?
a2 + b2 ≥ 2ab TRUE
x4 − 4x + 5 ≥ 0 TRUE
536x287196896 − 210y287196896 + 777x3y16z4732987
−1111x54987896 − 2823y927396 + 27x94572y9927z999
−936718x726896 + 887236y726896 − 9x24572y7827z13
+89790876x26896 + 30y26896 + 987x245y6z6876
+9823709709790790x28 − 1987y28 + 1467890461986x2y6z4
+80398600x2z12 − 27980186xy + 3789720156y2 + 9328769x
−1956820y − 275893249827098790768645846898z ≥ −389?
FALSE
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities
Question
Which of the following inequalities are true for all real valuesof the variables?
a2 + b2 ≥ 2ab TRUE
x4 − 4x + 5 ≥ 0 TRUE
536x287196896 − 210y287196896 + 777x3y16z4732987
−1111x54987896 − 2823y927396 + 27x94572y9927z999
−936718x726896 + 887236y726896 − 9x24572y7827z13
+89790876x26896 + 30y26896 + 987x245y6z6876
+9823709709790790x28 − 1987y28 + 1467890461986x2y6z4
+80398600x2z12 − 27980186xy + 3789720156y2 + 9328769x
−1956820y − 275893249827098790768645846898z ≥ −389?
FALSE
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities
Question
Which of the following inequalities are true for all real valuesof the variables?
a2 + b2 ≥ 2ab TRUE
x4 − 4x + 5 ≥ 0 TRUE
536x287196896 − 210y287196896 + 777x3y16z4732987
−1111x54987896 − 2823y927396 + 27x94572y9927z999
−936718x726896 + 887236y726896 − 9x24572y7827z13
+89790876x26896 + 30y26896 + 987x245y6z6876
+9823709709790790x28 − 1987y28 + 1467890461986x2y6z4
+80398600x2z12 − 27980186xy + 3789720156y2 + 9328769x
−1956820y − 275893249827098790768645846898z ≥ −389?
FALSE
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities, continued
Question
Can a computer decide, given a polynomial inequality
f (x1, . . . , xn) ≥ 0
with rational coefficients, whether it is true for all realnumbers x1, . . . , xn?
YES! (Tarski 1951) More generally, it can decide the truth ofany first-order sentence involving polynomial inequalities.
How? For example, how could it decide whether a given setdefined by a Boolean combination of inequalities is empty?
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Polynomial inequalities, continued
Question
Can a computer decide, given a polynomial inequality
f (x1, . . . , xn) ≥ 0
with rational coefficients, whether it is true for all realnumbers x1, . . . , xn?
YES! (Tarski 1951) More generally, it can decide the truth ofany first-order sentence involving polynomial inequalities.
How? For example, how could it decide whether a given setdefined by a Boolean combination of inequalities is empty?
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Inequalities: induction on the number of variables
x2 + y2 < 1
x > −1 ∧ x < 1
In general, the projection (x1, . . . , xn) 7→ (x1, . . . , xn−1)maps a set S defined by an explicit Booleancombination of inequalities to another such set S ′.
S 6= ∅ if and only if S ′ 6= ∅.Keep projecting until only 1 variable is left; then usecalculus.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Exponential inequalities
Can a computer decide the truth of inequalities like
eex+y
+ 20 ≥ 5x + 4y ?
Warmup: What about ee3/2
+ e5/3 ≥ 13396
143?
This should be easy: compute both sides to high precision,but. . .
What if they turn out to be exactly equal?
Schanuel’s conjecture in transcendental number theorypredicts that “coincidences” like these never occur, but ithas not been proved.
Theorem (Macintyre and Wilkie)
If Schanuel’s conjecture is true, then exponential inequalitiesin any number of variables are decidable.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Exponential inequalities
Can a computer decide the truth of inequalities like
eex+y
+ 20 ≥ 5x + 4y ?
Warmup: What about ee3/2
+ e5/3 ≥ 13396
143?
This should be easy: compute both sides to high precision,but. . .
What if they turn out to be exactly equal?
Schanuel’s conjecture in transcendental number theorypredicts that “coincidences” like these never occur, but ithas not been proved.
Theorem (Macintyre and Wilkie)
If Schanuel’s conjecture is true, then exponential inequalitiesin any number of variables are decidable.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Exponential inequalities
Can a computer decide the truth of inequalities like
eex+y
+ 20 ≥ 5x + 4y ?
Warmup: What about ee3/2
+ e5/3 ≥ 13396
143?
This should be easy: compute both sides to high precision,but. . .
What if they turn out to be exactly equal?
Schanuel’s conjecture in transcendental number theorypredicts that “coincidences” like these never occur, but ithas not been proved.
Theorem (Macintyre and Wilkie)
If Schanuel’s conjecture is true, then exponential inequalitiesin any number of variables are decidable.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Trigonometric inequalities
Question
Can a computer decide the truth of inequalities involvingexpressions built up from x and sin x?
NO! (Richardson 1968)
Idea: Let p, L ∈ Z[x1, . . . , xn] be such that L(~x)� p(~x)2.
f (~x) := −1 + 4p(~x)2 + L(~x)(sin2 πx1 + · · ·+ sin2 πxn).
If f (~x) < 0, then
sin2 πxi ≈ 0, so xi is very close to an integer ai , and
p(~x) < 1/2, which forces p(a1, . . . , an) = 0
Conclusion:
f < 0 somewhere ⇐⇒ p(~x) = 0 has an integer solution
(undecidable)
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Trigonometric inequalities
Question
Can a computer decide the truth of inequalities involvingexpressions built up from x and sin x?
NO! (Richardson 1968)
Idea: Let p, L ∈ Z[x1, . . . , xn] be such that L(~x)� p(~x)2.
f (~x) := −1 + 4p(~x)2 + L(~x)(sin2 πx1 + · · ·+ sin2 πxn).
If f (~x) < 0, then
sin2 πxi ≈ 0, so xi is very close to an integer ai , and
p(~x) < 1/2, which forces p(a1, . . . , an) = 0
Conclusion:
f < 0 somewhere ⇐⇒ p(~x) = 0 has an integer solution
(undecidable)
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Inequalities in one variable
Question
Can a computer at least decide the truth of trigonometricinequalities in one variable?
NO! In fact, the one-variable inequality problem is just ashard as the many-variable inequality problem.
The proof uses the parametrized curve
~G (t) := (t sin t, t sin t3).
What does this curve in R2 look like?
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Inequalities in one variable
Question
Can a computer at least decide the truth of trigonometricinequalities in one variable?
NO! In fact, the one-variable inequality problem is just ashard as the many-variable inequality problem.
The proof uses the parametrized curve
~G (t) := (t sin t, t sin t3).
What does this curve in R2 look like?
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
As t ranges over real numbers,
~G (t) := (t sin t, t sin t3)
traces out
For a continuous function f (x , y),
f (x , y) ≥ 0 on R2 ⇐⇒ f ( ~G (t)) ≥ 0 for all t
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
As t ranges over real numbers,
~G (t) := (t sin t, t sin t3)
traces out
For a continuous function f (x , y),
f (x , y) ≥ 0 on R2 ⇐⇒ f ( ~G (t)) ≥ 0 for all t
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Equality of functions
Bad news for automated homework checkers:
Theorem
It is impossible for a computer to decide,given two functions built out of x , sin x , | |,whether they are equal.
Proof: If you can’t decide whether f (x) ≥ 0, then you can’tdecide whether f (x) and |f (x)| are the same function.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Complex analysis
Example
Does
ez = w3 + 5z + 4
ew = w2 + 3z4 − 7
w4 = z9 + z5 + 2.
have a solution in complex numbers z and w?
Question
Can a computer decide whether a system of equationsinvolving the complex exponential function has a complexsolution?
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Complex analysis
Question
Can a computer decide whether a system of equationsinvolving the complex exponential function has a complexsolution?
NO! (Adler 1969)Proof: The 3 steps below characterize Z in C by equations:
1. 2πiZ is the set of solutions to ez = 1
2. Q ={a
b: a, b ∈ 2πiZ and b 6= 0
}3. Z is the set of q ∈ Q such that 2q ∈ Q; thus
Z := {q ∈ Q : ∃z ∈ C such that ez = 2 and eqz ∈ Q}.
Thus
Hilbert’s tenth problem ⊆ the complex analysis problem.
Hilbert’s tenth problem is undecidable,so the complex analysis problem is undecidable.
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Integration
Question
Can a computer, given an explicit function f (x),
1. decide whether there is a formula for∫f (x) dx ,
2. and if so, find it?
Theorem (Risch)
YES.
Theorem (Richardson)
NO.
Another answer: MAYBE; it’s not known yet.
All of these answers are correct!
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Integration
Question
Can a computer, given an explicit function f (x),
1. decide whether there is a formula for∫f (x) dx ,
2. and if so, find it?
Theorem (Risch)
YES.
Theorem (Richardson)
NO.
Another answer: MAYBE; it’s not known yet.
All of these answers are correct!
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Integration
Question
Can a computer, given an explicit function f (x),
1. decide whether there is a formula for∫f (x) dx ,
2. and if so, find it?
Theorem (Risch)
YES.
Theorem (Richardson)
NO.
Another answer: MAYBE; it’s not known yet.
All of these answers are correct!
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Integration
Question
Can a computer, given an explicit function f (x),
1. decide whether there is a formula for∫f (x) dx ,
2. and if so, find it?
Theorem (Risch)
YES.
Theorem (Richardson)
NO.
Another answer: MAYBE; it’s not known yet.
All of these answers are correct!
Undecidability ingroup theory,topology, and
analysis
Bjorn Poonen
Group theory
F.p. groups
Word problem
Markov properties
Topology
Fundamental group
Homeomorphismproblem
Manifold?
Knot theory
Analysis
Inequalities
Complex analysis
Integration
Integration
Question
Can a computer, given an explicit function f (x),
1. decide whether there is a formula for∫f (x) dx ,
2. and if so, find it?
Theorem (Risch)
YES.
Theorem (Richardson)
NO.
Another answer: MAYBE; it’s not known yet.
All of these answers are correct!